[EM] Automatic redistricting

[Anthony Simmons]

It might be difficult to base automatic districting on the
important considerations.  While we tend to think of
districts as geographical (perhaps because that's what they
utlimately are), social distinctions are just as important.
For example, we expect that districts will recognize
political boundaries.  We expect that they will (or will not)
group people with similar political concerns together.  These
social considerations would be difficult to quantify, or
perhaps even recognize.  I imagine it could be done, but I
wouldn't expect much agreement about whether it's been done
adequately.

Perhaps the trick is to let the voters create their own
districts, not necessarily geographical.  It might be better
to think of "consituencies".  After all, the concerns of
voters in modern democracies are not necessarily clumped
geographically.  Party lists assume instead that voters'
concerns aggregate by party.  But party is not necessarily
the dominant variable either.

The obvious way to allow voters to define their own
constituencies is with something like STV, which allows
voters to group themselves through the act of voting itself.
On the other hand, the effectiveness of this technique is
improved by using large multi-member districts, and it's been
pointed out that this increases demands on advertising, and
demands on the voters.

One solution would be to retain the idea of geographic
constituencies, but loosen the boundaries.  Instead of
putting a candidate on every ballot in a state, for example,
each candidate could be entered only on ballots within a
certain distance of her own location.  That distance would
have to be different for each candidate, so that all
candidates are on ballots for the same number of voters.

More precisely, a candidate's area (potential district, I'm
not sure what to call it, where that candidate appears on the
ballot) would extend different distances in different
directions, depending on population density in each
direction.  It would extend a shorter distance in directions
with high population density, and longer where the population
in that direction is sparser.

The idea would be to combine a system like STV, which allows
voters to define their own consituencies any way they like,
with the smaller ballots and other advantages of single-
member districts.  And this approach has the advantage of
being completely automatic; given good census data, a
computer could easily determine exactly which precincts would
be in each candidate's area, and in a very simple way that
would be easy for voters to accept.

    Tony


----- Original Message -----
Date: Fri, 22 Mar 2002 11:46:53 -0800
From: "Michael Rouse" <mrouse at cdsnet.net>
Subject: Re: [EM] More on Gerrymander prevention


The problem with sliver-shaped districts is definitely there. On the other
hand, compared to some of the gerrymandered monstrosities that certain
redistricting boards have come up with, this might be less of a problem than
it seems. I wish there was a program that would allow us to play with
various methods of apportionment and compactness -- one where you could
calculate medians, centroids, moments of inertia, perimeter measures, etc.
and apply them to real data -- that way you could determine how common such
problems would be and what the best method would be.

I definitely like the idea of automated districting though, one where there
is a clear and precise way of comparing two or more plans together and
chosing a winner. If there were some clear standard, we could (as others
have suggested) solicit plans not only from computer models and the parties
involved, but from ordinary citizens, and then choose the best one. Heck,
you could have a distributed-processing program like seti at home
(gerrymander at home? redistrict at home?) so that trillions of different options
could be tested (complete with cool screensavers) and the best ones
forwarded to the redistricting boards. If they were obligated by law to
choose the plan that scores the best by some standard method, computer geeks
(of which I'm a proud member) could keep the politicians honest.

Michael Rouse
mrouse at internetcds.com